Risk management in Treasury:

^{As part of the Corporate Treasury Newsletter, we have already looked at the use of AI in Treasury several times; related articles are linked accordingly.}

As the economic uncertainty increases, the relevance of risk management also increases. More specifically, efficient financial risk management can be a decisive factor in securing the continued existence of a company by anticipating financial risks and taking appropriate countermeasures.  

In a world that is almost completely globalized and where effects and counter-effects are intertwined in many ways, the dismissive saying “a sack of rice fell over in China” has its meaning being turned upside down in the real economy. This became evident, for example, when the “Ever Given”, a single container vessel, was wrecked and blocked the Suez Canal for six days, thereby severely impacting the global economy. While this example is an operational risk outside the focus of financial risk management, its ramifications nonetheless directly impact financial risk management. In this case, it directly affected the “classic” risks of financial risk management such as market risk, credit risk and liquidity risk. 

Advancing technologies, particularly artificial intelligence (AI), are opening up new opportunities to more quickly detect such correlations, making it easier to assess and manage the resulting risks. AI technologies are reshaping treasury and risk management in many ways and present companies with the opportunity to streamline processes and make them future-proof.

AI holds AI holds tremendous potential for financial risk management, particularly when it comes to the following areas: 

1. Identification of risks
AI is capable of analyzing massive amounts of data from various sources in order to identify risks faster and more precisely. Machine learning and data mining can be deployed to recognize patterns and anomalies in the data that indicate potential risks. Doing so results in more objective, faster and more accurate risk identification compared to traditional methods. This is particularly true for risks that are masked by complex correlations. 

• Credit risk: Artificial intelligence is able to assist in assessing the creditworthiness of business partners and customers by analyzing historical payment data, socio-demographic information and other relevant data.
• Market risk: AI models are capable of analyzing market trends and behavior to identify potential market risks such as price volatility and market liquidity.
• Fraud risk: Through analyzing transaction patterns, AI can flag unusual activity that indicates fraud or other unethical behavior.
• Operational risk: AI has the ability to pinpoint weak points in processes by spotting patterns in data that indicate inefficient or error-prone workflows.
• Liquidity risk: AI can analyze cash flows and liquidity requirements to forecast potential liquidity bottlenecks.

2. Assessing and predicting risk
One of the most significant benefits of AI in financial risk management lies in its ability to predict potential future risks and gauge the likelihood of their occurrence as well as the magnitude of the potential impact on a business. Predictive analytics takes historical data from various sources to forecast future events with potential risks. This allows companies to proactively take measures before the risk materializes. Based on the forecasts, measures for risk reduction can then be implemented and monitored in a more targeted manner. 

3. Automation and increased efficiency
AI solutions can also help boost efficiency in risk management by automating routine and time-consuming tasks. In this way, risk managers can focus on more complex and strategic risk aspects. By automating data collection and analysis, for example, risk managers can respond more quickly to changing conditions and make informed decisions.

4. Compliance and regulatory requirements
Compliance with statutory, regulatory and financing requirements is a key component of financial risk management. In this context, AI helps to minimize compliance risks by continuously monitoring compliance with regulations and automatically generating reports. In doing so, the risk of non-compliance and the associated financial and reputational damage is reduced.

Admittedly, integrating artificial intelligence into financial risk management offers considerable advantages, but it also brings with it specific challenges. In fact, these challenges are indicative of the limitations of AI.  

1. Data quality and availability
AI systems rely on large amounts of high-quality data to be effective. However, obtaining real-world data that is clean, well-structured, and comprehensive can be difficult. Incomplete or inaccurate data can lead to unreliable predictions and analyses. In other words, the use of AI models is constrained by the data that is available in sufficient quality. This is where the costs of improving quality and acquiring data should be weighed up against the potential benefits before deployment.

2. Model complexity and explainability
AI models, and especially those built using deep learning, can be extremely complex. Not only does this complexity make it difficult to understand how decisions are made, but it can also get in the way of maintaining and updating the models. Often, little insight can be gained into the decisions made by AI systems, making them difficult to explain. However, when it comes to risk management, it is crucial that decision-making processes are clear and reproducible. When the causes and characteristics of a risk cannot be explained, it becomes difficult to react in a targeted manner.  

3. Overfitting
AI models could become over-trained on the nuances of training data and as a result, perform poorly on new or changing data. This is especially problematic in financial risk management where market conditions can fluctuate rapidly.

4. Reliance on technology
An over-reliance on AI systems can lead to human experts being less involved in decision-making processes, with the risk of errors resulting from a lack of human judgment. In the case of risks that are subject to human irrationalities, this can be particularly problematic, owing to the limitations of the mechanisms used. Among these risks are, for example, geopolitical risks. 

5. Data protection and data ethics
Given that the analyses in financial risk management are always based on highly sensitive data, it is imperative to ensure that the data used is protected from unauthorized access at all times. Right from the consolidation and storage of the data used, it must be ensured that the IT systems and applications used meet the compliance requirements. Where personal data is incorporated into the AI in addition to the company's own data, it must also be ensured that the GDPR's requirements are met. On top of this, it is important to check whether the AI fulfills the ethical requirements of the EU's AI law and to what extent there is a reporting requirement. 

There is great potential for using AI in financial risk management, as the relevant risks are subject to certain logics and, thanks to modern treasury management systems, there is often a good data basis. That is why a good business case can be made for treasury to use AI. Still, artificial intelligence cannot replace the human risk manager – nor should it. Rather, its purpose should be to provide the risk manager with a tool that assists with routine tasks and developing and running through scenarios. On the one hand, by efficiently dividing the work between humans and AI, better results can be achieved. On the other hand, it is possible to consider correlations that were previously ignored due to time restrictions.

Source: KPMG Corporate Treasury News, Edition 147, September 2024
Authors:
Nils Bothe, Partner, Finance and Treasury Management, Corporate Treasury Advisory, KPMG AG
Tobias Riehle, Manager, Finance and Treasury Management, Corporate Treasury Advisory, KPMG AG